Learning With AI Is Hard

I’ve often thought that learning with AI is one of the best use cases for AI.

However, I think that it’s very common, or it will be very common, for people to run into two different problems with it.

The first is engagement. There are some data that suggests that critical thinking is being reduced by AI. And that’s not because of anything inherent with the AI tools, it’s just the passing off of our part in the critical thinking process to AI. So that part of the brain atrophies, as do lots of parts of the brain that actually don’t get used.

I think there is a need for you to really understand how engaged you are in the learning process, and if you feel yourself copying and pasting stuff and just entering and not engaging, you’re probably not learning.

There’s going to be some self‑monitoring involved with that.

But more importantly is a second issue, with how the learning is anchored.

I think learning works by attaching new ideas to what you already know. So Vygotsky called the sweet spot of this the zone of proximal development. So it’s close enough to our understanding to be reachable, far enough away to be genuinely new and worth learning.

The problem with using ChatGPT or Claude for learning is that every conversation essentially starts from zero. The AI doesn’t know what you already understand about a topic, what you’ve read, what frameworks you use, what your thinking has reached. So you have to spend part of every session teaching it those bits. Project instructions can help a bit, but they degrade over time and they’re static, which means that AI either defaults to too basic, which is stuff you already know, or too detached, it presents ideas that are of no connection to your existing mental models.

The shift happens when the AI can reference your existing thoughts during the conversation.

If it can read your notes, your sparks, your connections, the map of what you think and how you think, then it can do what a good teacher does; which is find the edge of your understanding and pitch the next idea just beyond it.

The biggest thing I’ve noticed with the Idea Factory is how much better I’ve got at learning because it knows exactly what I’m thinking at the time, what I’ve thought previously and what I should probably think about thinking next.